FFPP-Raw_1FPS_faces-expand-0-aligned_augmentation-normalize-image-mean-std
This model is a fine-tuned version of microsoft/swin-tiny-patch4-window7-224 on the imagefolder dataset. It achieves the following results on the evaluation set:
- Loss: 0.0034
- Accuracy: 0.9983
- Recall: 0.9929
- Precision: 0.9994
- F1: 0.9961
- Roc Auc: 1.0000
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
More information needed
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 5e-05
- train_batch_size: 16
- eval_batch_size: 16
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 64
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 20
Training results
Training Loss | Epoch | Step | Validation Loss | Accuracy | Recall | Precision | F1 | Roc Auc |
---|---|---|---|---|---|---|---|---|
0.1054 | 1.0 | 1377 | 0.0750 | 0.9716 | 0.9180 | 0.9495 | 0.9335 | 0.9957 |
0.0785 | 2.0 | 2755 | 0.0406 | 0.9853 | 0.9596 | 0.9723 | 0.9660 | 0.9986 |
0.0713 | 3.0 | 4132 | 0.0348 | 0.9878 | 0.9534 | 0.9899 | 0.9713 | 0.9994 |
0.0447 | 4.0 | 5510 | 0.0172 | 0.9933 | 0.9842 | 0.9851 | 0.9846 | 0.9997 |
0.0388 | 5.0 | 6887 | 0.0186 | 0.9936 | 0.9741 | 0.9964 | 0.9851 | 0.9998 |
0.0236 | 6.0 | 8265 | 0.0119 | 0.9957 | 0.9830 | 0.9971 | 0.9900 | 0.9999 |
0.031 | 7.0 | 9642 | 0.0137 | 0.9957 | 0.9928 | 0.9873 | 0.9900 | 0.9999 |
0.015 | 8.0 | 11020 | 0.0072 | 0.9972 | 0.9903 | 0.9969 | 0.9936 | 1.0000 |
0.0429 | 9.0 | 12397 | 0.0087 | 0.9967 | 0.9863 | 0.9987 | 0.9925 | 0.9999 |
0.0186 | 10.0 | 13775 | 0.0052 | 0.9979 | 0.9919 | 0.9985 | 0.9952 | 1.0000 |
0.0282 | 11.0 | 15152 | 0.0069 | 0.9974 | 0.9892 | 0.9988 | 0.9940 | 1.0000 |
0.0034 | 12.0 | 16530 | 0.0045 | 0.9979 | 0.9947 | 0.9956 | 0.9951 | 1.0000 |
0.0187 | 13.0 | 17907 | 0.0070 | 0.9972 | 0.9886 | 0.9986 | 0.9935 | 1.0000 |
0.0136 | 14.0 | 19285 | 0.0038 | 0.9982 | 0.9931 | 0.9988 | 0.9959 | 1.0000 |
0.006 | 15.0 | 20662 | 0.0039 | 0.9982 | 0.9928 | 0.9988 | 0.9958 | 1.0000 |
0.0067 | 16.0 | 22040 | 0.0037 | 0.9983 | 0.9926 | 0.9995 | 0.9960 | 1.0000 |
0.0121 | 17.0 | 23417 | 0.0036 | 0.9983 | 0.9929 | 0.9992 | 0.9960 | 1.0000 |
0.0026 | 18.0 | 24795 | 0.0037 | 0.9982 | 0.9925 | 0.9993 | 0.9959 | 1.0000 |
0.0024 | 19.0 | 26172 | 0.0034 | 0.9983 | 0.9932 | 0.9991 | 0.9961 | 1.0000 |
0.002 | 19.99 | 27540 | 0.0034 | 0.9983 | 0.9929 | 0.9994 | 0.9961 | 1.0000 |
Framework versions
- Transformers 4.39.2
- Pytorch 2.2.2
- Datasets 2.18.0
- Tokenizers 0.15.2
- Downloads last month
- 127
Finetuned from
Evaluation results
- Accuracy on imagefoldertest set self-reported0.998
- Recall on imagefoldertest set self-reported0.993
- Precision on imagefoldertest set self-reported0.999
- F1 on imagefoldertest set self-reported0.996